What is Name Screening?
Name screening is the process of comparing the names of customers, beneficial owners, and counterparties against sanctions, PEP, and other watchlists to find potential matches.
Name screening is the matching engine behind sanctions and watchlist screening. It decides whether a name in a firm's records is close enough to a name on a list to need human review. Because names are spelled, translated, and recorded in many different ways, getting the matching logic right is one of the hardest parts of screening.
Name screening must balance two risks. Matching too strictly can miss real sanctioned parties because of a small spelling difference. Matching too loosely floods analysts with false positives from people who simply share a common name.
How does name screening work?
Name screening uses exact matching and fuzzy matching together. Exact matching only flags names that are identical, which can miss variations like Muhammad and Mohamed. Fuzzy matching uses algorithms to find names that are similar, including the following.
- Levenshtein distance, which counts the edits needed to turn one name into another
- Jaro-Winkler similarity, which scores character matches and works well for short strings like names
- Phonetic algorithms such as Soundex and Metaphone, which group names that sound alike, such as Smith and Smyth
Firms set a minimum similarity score that triggers an alert, such as a Jaro-Winkler score of 0.85.
How do firms improve name screening accuracy?
Firms improve name screening accuracy by adding context and handling name variations. Common techniques include the following.
- Confirming matches with other details, such as date of birth, nationality, location, or passport number
- Checking every known alias listed for a sanctioned party
- Maintaining alias tables for common transliteration variants
- Recognizing names with multiple given or family names, such as Juan Carlos Fernandez and Juan C. Fernandez
- Applying stricter criteria to very common names unless other risk factors exist
A layered approach usually works best. Firms start with exact matches on full name plus date of birth or ID number, then apply fuzzy matching with contextual filters. Confirmed false positives can be whitelisted with a documented reason and reviewed periodically to make sure the person's status has not changed.
